Chapter 61
Obesity and Diabetes — 20
What do GPT-4 and AlphaFold2 have to do with GLP-1 drugs? Figure 4.5 provides a schematic of the structure of the two prototypic GLP-1 drugs, semaglutide and tirzepatide, and the biologic gut peptides (incretins) they activate. They look… complicated. Yet the new AI models can readily predict the three-dimensional structure resulting from any changes to a GLP-1 peptide drug.
With appropriate learning from all prior GLP-1 work and related chemistry and structural biology, the transformer AI models can design new peptide structures that would be anticipated to be more potent than existing molecules. This can even extend to small molecules—pills—that haven’t yet been conceived. We’ve already seen how AI led to a new structural class of antibiotics. The last time that occurred, it took more than three decades! This included discovery of molecules active against highly resistant bacteria, such as methicillin-resistant staphylococcus aureus. So had AI been available at the dawn of GLP-1 drugs getting approved more than twenty years ago, the path to longer half-life agents with more potent effects would likely have been much faster.
Figure 4.5. Glucagon-like peptide structures, natural (GLP-1 and GIP) and drugs (semaglutide and tirzepatide). Adapted from Eric Topol, “The new obesity breakthrough drugs,” Substack, December 10, 2022, https://erictopol.substack.com/p/the-new-obesity-breakthrough-drugs.
It is essential to recognize that there will be an acceleration of GLP-1 drugs, such as small molecules that can cross the blood-brain barrier with ease, providing stronger modulation on behaviors, or new molecules with substantially fewer gastrointestinal side effects. Like the plane that left late, AI is going to make up time in the future refining GLP-1 drugs as we know them today.
The impact of GLP-1 drugs in medicine—for improving patient outcomes—will likely be more substantial than any prior class of medications, including statins. The development of these drugs embodies a major triumph for the unmet need of high efficacy and safety for the huge population of people with obesity and type 2 diabetes. But the AI progress we have witnessed will help us with similar conquests over the greatest challenges.
We are now seeing biomedical discoveries at a breakneck pace, and they will be moving along at high velocity. Now that we have begun to crack obesity, one of the most daunting conditions, the future is bright for finding new solutions to many refractory killer diseases.